HP 20b Link 1

Link to HP 20b page on HP's website (if this works)

My detailed review of the HP 20b should be posted to the HPCC.org website (and in the next issue of Datafile) shortly.

hpcc.org

Highlights:

RPN, algebraic and chain logic modes.

Fast. I mean incredibly fast. Uses a 30Mhz CPU.

Extensive financial and statistical and mathematical functions...such as four probability distributions and inverses, trig and inverse trig in degrees or radians, hyperbolics and inverses, six regression models, standard error of the mean, covariance, discounted payback, Bonds, NFV, NUS, in addition to the "usual" financial functions.

12 digit mantissa display plus 3 digit exponent. Top line of display labels results.

What's not present? No programming. No HP solve.

30 learning modules should be up on the HP site soon.

The review probably contains answers to most questions, but I'll answer any posed as I can.

What is the funny dot type display in the top right corner in the brochure?

An Atmel AT91SAM7L128 ARM7 CPU - interesting!

That CPU is about 0.5mA/MHz, so at 30MHz that sucker is going to need 15mA from the two CR2032 batteries - hmm... somehow I don't think it's going to be doing much at 30MHz, if anything. I think it's 30MHz "capable" for marketing purposes, I'd be surprised if it actually runs at that speed, even for brief periods.

I like the girl in the brochure ;-)

Dave.

The top right corner contains annunciators for things like begin mode, 360-day mode, radians mode, etc.

Attempting to measure current drain using classical static methods and expressing it in static terms doesn't provide a realistic number. Digital circuits turn on and off so fast these days that unless you have a repetitive multi second operation, very carefully specified, the milliampere "number" won't be very meaningful.

Giving a "number" can only be very approximate, and very much limited to the carefully specified conditions of use. This becomes even more difficult as the speed (clock, and clocking phase type) increases.

Even an inactive, sleep, or hibernation mode is difficult to measure accurately. A good technique is to use a series resister and a millivolt (or microvolt) meter because switching ranges with a normal series meter traumatizes the calculator.

Of course a digital oscilloscope provides the most accurate time based current "measurement." Even the designers of these circuits have great difficulty - the difficulty is the amount of engineering time, not the methods - in making and specifying current drain values.

X < > Y,

Richard

Lack of programming is a serious bummer, but if the mechanical quality and mathematical precision is as good as even the current HP-12C, it's a winner IMO.

Going by the low-res pictures, it looks like they used a 7-segment numerical display. If so, I applaud them: it should be much more readable than dot matrix.

Any word on price?

Google found this page:

Connecting Your World Berlin 2008

There are high resolution images and data sheets for the HP20b, OfficeCalc 100 and 200, PrintCal 100, and Quick Calc.

Further a data streaming sensor interface for HP37gs and 40gs is presented, the HP StreamSmart 400.

The prices are already listed in HP online store

The prices in Germany (Europe?) were anounce in a press note available at pressebox.de

Again I am not amused about HP's price policy for Europe.

Hubert

Edited: 10 June 2008, 4:13 p.m.

If you click the "buying options" link, it takes you to a page "buspurchase-discontinued.html"!

Obviously it isn't discontinued, but it seems like they might want to have a page with a different URL. :-)

I did a Google search for "hp 20b", and found it in stock for $14, but I was a little skeptical so I didn't order it.

Edited: 10 June 2008, 2:23 p.m.

HPCC website where 20b review can be found

HP is having a small problem of running out of product numbers, for $14 you get a printer cartridge. Chuck

What you found is probably this

http://www.shopping.hp.com/product/C6614D?landing=supplies&category=ink_toner&family_name=

Quote:

---

Attempting to measure current drain using classical static methods and expressing it in static terms doesn't provide a realistic number. Digital circuits turn on and off so fast these days that unless you have a repetitive multi second operation, very carefully specified, the milliampere "number" won't be very meaningful.

Giving a "number" can only be very approximate, and very much limited to the carefully specified conditions of use. This becomes even more difficult as the speed (clock, and clocking phase type) increases.

Even an inactive, sleep, or hibernation mode is difficult to measure accurately. A good technique is to use a series resister and a millivolt (or microvolt) meter because switching ranges with a normal series meter traumatizes the calculator.

Of course a digital oscilloscope provides the most accurate time based current "measurement." Even the designers of these circuits have great difficulty - the difficulty is the amount of engineering time, not the methods - in making and specifying current drain values.

X < > Y,

Richard

---

Yep, I realise all that, this calc uses dynamic clock switching.
But if this device ups the clock rate to 30MHz briefly to do every calculation then the CR2032 batteries aren't going to last very long at all. Probably not the claimed "average of 9 months".

I suspect marketing have got hold of the 30MHz figure and think that sounds just terrific.

My uWatch operates at 250KHz and all calcs are near instantaneous on its little non-optimised 16bit micro. The 20B uses a 32bit ARM7 so I doubt it needs to run at the claimed 30MHz. Also, it's easy to choose the clock rate based on the function being performed.

Power consumption is not hard to calculate in the design phase BTW. You know the mA/MHz figure (which is reasonably accurate average figure), and you know how long it runs for at each clock rate. e.g. 5% of the time is spent doing calcs at XXMHz, 90 of the time is spent idling at XXKHz. So it's easy calculate a devices average current consumption.

This ATMEL ARM is a bit more complex as it has multiple clock domains, but you can still work out average figures. If this thing runs at 30MHz it's going to take 15mA on average @ 1.8V, simple as that.

Dave.

Has anyone noticed the AT91SAM7L128 micro used has 128KB of internal FLASH program memory.

There might be the potential here to reprogram the unit to customise it, as was recently discussed about the 35S.

Anyone opened one?

Also when HP say "HP’s latest high-speed calculator processor" that implies they intend to use this processor in other products.

Dave.

Edited: 10 June 2008, 6:23 p.m.

How low can they go? The Quick Calc nullifies 35 years of HP engineering. "Unique horizontal shape", "% calculations". That would have been nice features in, uhm, 1971.

Quote:

---

Again I am not amused about HP's price policy for Europe.

---

Edited: 10 June 2008, 5:56 p.m.

Gene, thanks a lot for making your review article accessible right now! You wrote they use blue ANS for LASTx in RPN mode. What will be the use of "=" then?

Now that the use of the AT91SAM7L128 in an HP product has been revealed publicly, I hope that I can say this without violating any NDA.

There was a very useful and interesting discussion at HHC2007 about power consumption with this part, especially with regards to circuits powering up only when needed, and shutting off when not needed. It's an area where the AT91SAM7L128 excels, and the power consumption per calculation should be better than in older products. Cyrille was really able to dig into some nitty-gritty technical details, and it became one of my favorite sessions from HHC. I'm really excited to see this come into production, and eager to get my hands one one and play with it!

There's always a few people who will slam on HP no matter what they do [I'm not pointing fingers or naming any names, but I'll bet we'll see some pop up in this thread before too long! ;) ], but I'm very pleased to see this calculator. I'll withhold judgment on build quality until I get it into my hands, but it certainly looks good to me. It really recalls the style of the Pioneer case, while still being updated and modern. I know that lack of programming may be seen as a bad thing, but not *every* product needs to be or should be programmable, I think. I have absolutely no doubts about HP's commitment to programmability on other product lines. So good on you, HP! Thanks!

= serves as ENTER in RPN mode. = and INPUT will perform the same function in "run" mode.

In menus, they operate somewhat differently. = will compute a result if possible in a menu.

Certainly looks like a promising option....

- Pauli

That's a pity. I'd have welcomed the use of INPUT for ENTER in RPN (or vice versa ;) But can be changed...

Quote:

---

if this device ups the clock rate to 30MHz briefly to do every calculation then the CR2032 batteries aren't going to last very long at all

---

That's most likely incorrect. Within reasonable frequency bounds, for performing a fixed amount of computation, CMOS circuits generally consume approximately the same energy regardless of clock rate. At a higher clock rate, it will draw more power, but for a proportionally shorter interval.

If you tell it to compute an interest rate, for example, it only has to do a fixed amount of computation, and it will consume about the same energy to do that computation whether it is running at 1 MHz or 15 MHz.

The exceptions primarily occur when there are side effects to the higher clock speed. For instance, if the flash only supplied 32-bit data and could only run at 15 MHz, running the CPU at 30 MHz would just introduce wait states, so the computation would consume more energy that it would at 15 MHz.

As it happens, the flash in the AT91SAM7L128 actually is limited to 15 MHz. However, it is 128 bits wide (four instructions). Since the majority of instruction fetches are consecutive, the CPU can run at 30 MHz without incurring a wait state on every instruction. This means that running it at 30 MHz will be somewhat less energy-efficient than at 15 MHz, but somewhere above 1/2.

So much for sardonic comments. :-)

At least, it would have fit in Mr. Hewlett's shirt pocket, something the OfficeCalcs seem not to be able to.

:-)

Quote:

---

That's most likely incorrect. Within reasonable frequency bounds, for performing a fixed amount of computation, CMOS circuits generally consume approximately the same energy regardless of clock rate. At a higher clock rate, it will draw more power, but for a proportionally shorter interval.

---

That's true, but there are other practical considerations in designs like this. Such as the losses in the internal battery resistance, which is substantial with CR2032's, in the order of 20-40ohms over the operational range.

It's that I^R factor that gets you.

So let's say you operate at 1MHz for 10 seconds @ 0.5mA with a 20ohm battery resistance.

That's 5uW for 10 seconds, or 50uW/second loss in the battery.

Increase the clock rate to 10MHz and you can do the same stuff in 1 second instead of 10 seconds. That's 5mA for 1 second which is 500uW for 1 second, or 500uW/seconds loss in the battery.

Back of envelope figures of course.

Bingo, you've just wasted 10 times the energy in the battery losses to perform the same function at 10MHz as opposed to 1MHz.

Dave.

... by some very rapid prototyping, please find below an as well preliminary scientific version of this new business calculator:

It can be done d8)

Quote:

---

Now that the use of the AT91SAM7L128 in an HP product has been revealed publicly, I hope that I can say this without violating any NDA.

There was a very useful and interesting discussion at HHC2007 about power consumption with this part, especially with regards to circuits powering up only when needed, and shutting off when not needed. It's an area where the AT91SAM7L128 excels, and the power consumption per calculation should be better than in older products. Cyrille was really able to dig into some nitty-gritty technical details, and it became one of my favorite sessions from HHC. I'm really excited to see this come into production, and eager to get my hands one one and play with it!

There's always a few people who will slam on HP no matter what they do [I'm not pointing fingers or naming any names, but I'll bet we'll see some pop up in this thread before too long! ;) ], but I'm very pleased to see this calculator. I'll withhold judgment on build quality until I get it into my hands, but it certainly looks good to me. It really recalls the style of the Pioneer case, while still being updated and modern. I know that lack of programming may be seen as a bad thing, but not *every* product needs to be or should be programmable, I think. I have absolutely no doubts about HP's commitment to programmability on other product lines. So good on you, HP! Thanks!

---

Thanks for that Seth. The new Atmel ARM processor does look very flexible in the dynamic power consumption area, an excellent choice to base future calcs on I think.

I think this calculator really is a BIG DEAL. It likely points in the direction HP are going to go. i.e. a fast FLASH ARM processor even in low cost units (this is a $40 retail calc), dual 7 segment/dot matrix display, and that lovely looking modern Pioneer-like case styling with the nice long rubber feet on the bottom. I really like it.

My only complaint would be the battery life. 9 months doesn't cut the mustard IMHO. They should have used AAA's.

Dave.

INPUT does work as ENTER in RPN mode.

No worries. Just ignore that = key if you like.

Gene

Quote:

---

... by some very rapid prototyping, please find below an as well preliminary scientific version of this new business calculator:

It can be done d8)

---

Very nice, as usual Water.
But the lack of a dedicated EEX key would be inexcusable!

With so few keys, such a calc would have to have its key layout optimised for non-programability.

Dave.

So "=" is a useless key in RPN mode on the 20b??

Quote:

---

Very nice, as usual Water. But the lack of a dedicated EEX key would be inexcusable!

---

Quote:

---

With so few keys, such a calc would have to have its key layout optimised for non-programability.

---

I am not convinced.

I won't like the tactile feeling of the keyboard, so I won't use it. Again HP will disappoint many of us.

Quote:

---

I am not convinced.

I won't like the tactile feeling of the keyboard, so I won't use it. Again HP will disappoint many of us.

---

Geeze, try some
optimism!

Kinda reminds me of people who buy $1000 Monster brand speaker cable for their stereo system. Absolutely no difference to a $10 one, but they have already convinced themselves otherwise, and so it does.

Dave.

= and INPUT will both function as ENTER in RPN mode. This is exactly the same as the 17b2 and 17b2+, so this is nothing new.

What is new is the use of = to compute results while in a menu environment.

hello,

actually, no it can not, the CPU only supports a 400 pixel screen, and your screen design has way more than that.... so sadly enough, it is not doable with that platform.

sorry, cyrille

Gene writes in his review:

The 20b keyboard is not the 50g or 35s keyboard. Pressing a key does not give that good old “click” that you hear on those models. That said, it’s not a squishy keyboard by any means either, as some competitors use, where you can’t tell if the key has been pressed or not. It is somewhere in the middle. ... It is certainly better than the competitor’s keyboards.

In comparision to the pre 1999 calculators, I already have a problem with the 35s keyboard. So it seems that this new keyboard does not bring any improvement to me. I want the keyboards we had with the 48, 42, 32, 41 etc.

As long as HP does not give me that feature, I have to stick with those older calculators and won't use the newer ones.

Less power, but a decent hardware has more value for me than power with a bad hardware.

Thanks for the review, Gene. I think this will replace my 17bII+. I rarely use the Solver, and I could use the additional math functions.

Aside from the Solver, alarm clock, and IR printing, what does the 17bII+ have that the 20b does not have? The 20b seems to be pretty well stocked with financial functions.

hello,

le me try to answer a bunch of things at once, to make it easier...

- agregated questions about CPU speed and battery issues:
Yes, it is running at 30Mhz while doing calculations. With few exceptions, calculations are done in less than 100ms (even IRR type stuff, the longest calculations are, surprisingly enough, probability distribution ones)... the battery drain is large at these frequency (15mA or so), and that is why we are using 2 battery. since 1 CR2032 is rated for a pulse current of 20mA for 15s, this is in theory completely OK...
As soon as calculations are finished, it pretty much shuts down (ie, all gets powered off with the exception of the screen).

- question about 15Mhz flash speed:
Flash speed is 15mhz and CPU is 30Mhz, BUT most of the code is in thumb mode where instructions are only 16 bits, so thumb code does execute at the full 30Mhz. some of the critical loop code is in ARM mode, but that is also OK thanks to the buffer and small cache that allows that code to also run at full speed even at 30Mhz.

- My only complaint would be the battery life. 9 months doesn't cut the mustard IMHO. They should have used AAA's.

the calculator is the thickness of a AAA! using AAA would have made the unit significanly bulkier and not as 'pretty'...

- losses in the internal battery resistance, (in the order of 20-40ohms):
datasheets indicate <10ohms (typically 7ohms)... you of course need to divide this by 2 since there are 2 battery in parallel, so typically 3.5 Ohms...
as a result, although, there is still a *10 of the power consumption in the battery at higher speed, that consumption is negligeable compared with the CPU consumption.

- Going by the low-res pictures, it looks like they used a 7-segment numerical display. If so, I applaud them: it should be much more readable than dot matrix.
there is both a large dot matrix (ie: large dox, small matrix) on the top line and a 7 segment display at the bottom for full 12+3 accuracy display (12 digits, 3 exponents).

- Again I am not amused about HP's price policy for Europe.
it's not the HP price policy. remember, HP is not allowed to 'fix' prices, that would be illegal (and conveniently called... 'price fixing':-). also remember that you will pay an extra 20% or so in europe due to VAT while the US price are always given with no tax...

- Has anyone noticed the AT91SAM7L128 micro used has 128KB of internal FLASH program memory (well, yes, I have!)
There might be the potential here to reprogram the unit to customise it, as was recently discussed about the 35S.
yes, there is a potential, and conveniently enough, the poor engineers have inadvertently left the footprint for the JTAG port on the final PCB, so you can easely open the calculator, connect a JTAG and go hacking happily! But I guess we will find someone in the group to complain about the 'lack of finish' and HP shipping stuff that is 'barely more than prototypes' because of that!

- Anyone opened one?
nope, I have opened 100's... but that is just me bragging :-)

regards, cyrille

Hi Cyrille,

Quote:

---

the CPU only supports a 400 pixel screen

---

On the other hand, didn't the 27S and 42S have LCDs with 16x131 dots in 1988? And are you telling me now we cannot have a calc with twice this display size (or resolution) 20 years later?? Looking at all the intriguing gadgets available off the shelf (almost) everywhere with a lot of details on their versatile displays you see me quite a bit confused. OK, they cost a bit more than 40 US$, but I'd be willing to lay down a bit more than this for a decent device. And I'm not the only one for sure.

Presumably I missed the point? Please explain to an old guy like to a 4-year-old, else I'll have no chance.

Thanks in advance,

Walter

Edited: 11 June 2008, 10:18 a.m.

Quote:

---

Less power, but a decent hardware has more value for me than power with a bad hardware.

---

Power is nothing without control
[Pirelli Tyres]

;-)

hello,

can I explain as if you were a 40 year old :-)

in order to keep cost down, it is important to find an integrated chip with a build in screen driver. if you do not do that, you end up with a multi chip solution (main CPU and separated screen drivers like on the 50g).

strangely enough, they are plenty of 8 bit chips with screen drivers, up to size of 6 or 7000 pixels, but there is no other commercially available ARM chip that I know of with a screen driver!

it looks like, in the mind of chips manufacturers, there is no reason to use a 32 bit chip in hand held electronics... so no-one is making ARM based screen driving chips!

this is why, at this point in time, if you want to use the current platform, you are limited to 400 pixels (to be precise, a 6*43 dot matrix plus 11 indicators + 131 pixels for the 'number display')...

in order to use a screen with more pixels, we would need to use an external screen driver, or change CPU, which is outside of the scope of that project.

so, if you can do with the current 400 pixel screen, you will be able to reprogram the calculator flash with your software and have a 20b transfromed into your own WB 42s... but if you want a larger screen, you would need to redo the electronics...

cyrille

HiGene,

In your review, you wrote "A larger PDF manual is available online
as well as 30 learning modules that go into greater detail on how to use the 20b."

where can these be found at hp's website?

Thanks

hpnut in Malaysia

These aren't there yet. At the time of the review, I had hoped these would be online shortly. I have no idea when they will actually get put online and no, I can't share them "unofficially". To whet your appetites, however, here is the list of the topics of the 30 learning modules:

Introduction to the learning Modules.doc

Using RPN and the stack.doc

Algebraic and ChainMode.doc

Basic Arithmetic.doc

Operating Modes and Error Messages.doc

Using Memories.doc

Date calculations.doc

Percents.doc

Markup, Margin, PercentChange.doc

Time value of Money Basics.doc

Simple and Compound Interest.doc

Loan Payments.doc

Down Payments.doc

Saving for Retirement.doc

Interest rate Conversions.doc

Balloon Payments.doc

Lease Payments.doc

NPV and other DCF measures.doc

IRR.doc

Amortizations.doc

Depreciation.doc

Bonds.doc

Breakeven.doc

Rearranging Items.doc

Averages Std Deviations.doc

Regression.doc

Probability Distributions.doc

Logs, Exponentials, and other Math.doc

Trigonometry.doc

Hyperbolics.doc

Those keyboards are gone and unlikely to ever return, so you may be stuck using the older machines.

That said, the 50g keyboard is quite nice. I don't know why that keyboard wasn't used in the 20b, but my suspicion is price. At a rough $40 retail price, there isn't much room for possibly expensive engineering. Tough call.

And, to repeat, I have NEVER had a missed keystroke on the 20b. I just spent 5 minutes TRYING to cause a double click or a missed click and I cannot get it to happen.

The 20b is very attractive if you can live without the solver or if you really would use some of the added statistical and mathematical functions.

The solver is a very big deal for some of us. :-)

Quote:

---

strangely enough, they are plenty of 8 bit chips with screen drivers, up to size of 6 or 7000 pixels, but there is no other commercially available ARM chip that I know of with a screen driver!

---

And some of those 8-bit micros cost less than $0.50 in high volume, as do the standalone display driver chips used in graphic LCD modules. The obvious solution is to use an ARM for the main processor, and either a separate LCD driver or a separate microcontroller with integral LCD drive.

Sure it costs another $0.50, but when used in a relatively more powerful calculator, it can also be sold for $5 or $10 more.

Cyrille, Gene, Eric,

if I understand you correctly (thanks Cyrille, I'm feeling more than a decade younger now at least), then within the range of recent RPN calculators

• a primary "=" is not necessary for operation,
  
• additional material cost of 0.5US$ would qualify the HW to drive a 132x48 dot matrix display.
  

Though total additional manufacturing cost will for sure exceed these 0,5US$, I think it being worthwile, since such a bigger display will allow for

1. a line of soft keys making use of the full display width, being accessible most intuitively and fastest via the keys located next to this line,
   
2. displaying everything one may think of within this dot matrix, including (but not limited to) status indicators of any kind, multiple lines of results, matrices, equations etc.
   
3. giving the device a friendlier, more open and modern appearance instead of looking to you like through the arrow slit of a medieval black fortress.
   

As usual, my 20 Milli-Euros only

Quote:

---

- My only complaint would be the battery life. 9 months doesn't cut the mustard IMHO. They should have used AAA's.

the calculator is the thickness of a AAA! using AAA would have made the unit significanly bulkier and not as 'pretty'...

---

Yep, fair enough.
How about a thicker coin cell then? Although I know 2032 is probably the most popular and lowest cost option.

Or how about dropping that clock rate?, does it *really* need to be 30MHz?

Quote:

---

- losses in the internal battery resistance, (in the order of 20-40ohms):
datasheets indicate <10ohms (typically 7ohms)... you of course need to divide this by 2 since there are 2 battery in parallel, so typically 3.5 Ohms...

---

What battery are you using?

An Energizer CR2032 has at least 15 ohms internal resistance when fresh, rising to 40 ohms or so by the end of it's useful life:

http://data.energizer.com/PDFs/cr2032.pdf

Sony also spec about 20ohms (but you have to do some math to get it).

Many brands don't specify it unfortunately.

I am aware of the rule of thumb in the industry for 2032's being 20 ohms.

Quote:

---

as a result, although, there is still a *10 of the power consumption in the battery at higher speed, that consumption is negligeable compared with the CPU consumption.

---

Using your figure of 3.5ohms that's 0.78mW being wasted every operation, with the processor (and linear LDO) taking roughly 42mW (2.8V*15mA). So that's 1.8% using your best case figures - ok , fairly insignificant. However, it would increase significantly as the battery life drops, or potentially very significantly using cells other than the one you specify, such as the popular Energizer.

Quote:

---

- Has anyone noticed the AT91SAM7L128 micro used has 128KB of internal FLASH program memory (well, yes, I have!)
There might be the potential here to reprogram the unit to customise it, as was recently discussed about the 35S.
yes, there is a potential, and conveniently enough, the poor engineers have inadvertently left the footprint for the JTAG port on the final PCB, so you can easely open the calculator, connect a JTAG and go hacking happily!

---

NICE!, VERY NICE!

So very nice in fact that it's cause for celebration!

Dave.

Edited: 11 June 2008, 10:09 p.m.

Of course, some people (like myself) prefer a 7 segment main display.
I'm a big fan of dual 7 segment and dot matrix displays.

But I do understand this is inadequate for a high end fully programmable 42S-like calc with soft keys.

I'd be MASSIVELY happy with a basic scientific version of the 20B with the current display!

Oh, and an extra row of dedicated keys would not go astray of course.

Dave.

Quote:

---

- Anyone opened one?
nope, I have opened 100's... but that is just me bragging :-)

---

Photo?

Is the CPU the QFP or the BGA version? On the former we can get to any extra signal lines and cut things more easily.....

- Pauli

Speaking of trigonometry. . . how about the "ever present" (hopefully not ever again) COS bug?

TW

Edited: 12 June 2008, 12:06 a.m.

From the 35s bug list article:

35s bug list article

"For instance, cos(89.999) returns 1.745329091E-5. The correct result at this precision is 1.74532925191E-5."

On the HP 20b, COS(89.999) returns 1.74532925191E-5.

No bug!

Hmm. Wonder what those six holes are for? ;)

Hi, Tim and Gene --

It would seem that the bug has been addressed, but IMNSHO the definitive test is found here:

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv017.cgi?read=123880#123880

Note that results with 10 or more input digits were correct on the HP-33s and HP-35s with flawed trig functions.

Only if you can duplicate the complete HP-32SII results on the new HP-20b, would I say conclusively that the bug has been exterminated.

-- KS

Edited: 12 June 2008, 1:16 a.m.

I get the EXACT results that the 32SII gets with this slight difference:

                 SIN - 32SII          SIN - 20b            ULP Diff

3.1415926        5.35897932384E-08    5.35897932385E-08       1

All other results on the 20b from Karl's link are identical to the 32SII results.

Is the last digit truly supposed to be a 4 or a 5?

Is the 20b using more digits than the 32SII internally and might that cause this one result to be different?

hello,

answers to 2 questions:

- the 6 holes are to connect a serial cable (either for custom application, or for reprogramming the calculator)

- the CPU is in die form, so you can not access it,however, strangely enough, the same 'bad' engineers we talked about earlier also 'forgot' to remove 2 10 pin foot print that connects respectively to 8 digital IO and the 4 ADC channels...

regards, cyrille

Quote:

---

hello,

answers to 2 questions:

- the 6 holes are to connect a serial cable (either for custom application, or for reprogramming the calculator)

- the CPU is in die form, so you can not access it,however, strangely enough, the same 'bad' engineers we talked about earlier also 'forgot' to remove 2 10 pin foot print that connects respectively to 8 digital IO and the 4 ADC channels...

---

Absolutely frigg'n brilliant!

Dave.

The 15 digit result is 5.35897932384626E-08

If they really are using more digits, you'd think there would be more than just that one difference.

Quote:

---

If they really are using more digits, you'd think there would be more than just that one difference.

---

A question:

I love the fixed exponent display, but the display shows the mantissa on the left hand side. Does that mean an exponent display will be:

"1.23_________-3" ?

Or does the mantissa right shift when an exponent is displayed?

I've never liked displays with a left justified mantissa and a right justified exponent, with all that white space, it's just wrong IMHO.

Dave.

Edited: 12 June 2008, 9:41 a.m.

Quote:

---

Going by the low-res pictures, it looks like they used a 7-segment numerical display. If so, I applaud them: it should be much more readable than dot matrix.

---

From the pictures in the Gene's review, it appears to have a 7-segment display on the bottom line, and dot matrix on the top line. Is this possible? If so, it seems best of both worlds: segmented for numerals, matrix for alpha.

Yes, it displays the mantissa on the left side of the display and the exponent on the right side of the display.

That's exactly how the HP35, HP45, HP55, HP65, HP67....even the HP12c do it.

So, consistency! :-)

The exponent is smaller and slightly raised as well.

There are 3 digits available for the smaller exponent. If the exponent requires only one digit, then the 1 digit is left-justified in this 3-digit space. If it requires 2 digits, it is left-justified. A small minus sign is present between the last digit of the mantissa and the exponent field. It is also slightly raised to be level with the exponent.

Gene --

Thanks for executing the tests on the HP-20b.

                 SIN - 32SII          SIN - 20b            ULP Diff

3.1415926        5.35897932384E-08    5.35897932385E-08       1

The fault was my own. Indeed, the HP-32SII (and, I'm sure, any Pioneer-series calc) produces the correctly-rounded 5.35897932385E-08 as the 12-digit result. As you might expect, I did a lot of copying and pasting to prepare that table, and made an editing error.

OK, I'm convinced that the trigonometric computational bug was successfully addressed. In case HP had not already investigated the root cause of the bug, I provided the URL link containing my tables to Cyrille at the 2007 HHC.

-- KS

Quote:

---

- the CPU is in die form, so you can not access it,however, strangely enough, the same 'bad' engineers we talked about earlier also 'forgot' to remove 2 10 pin foot print that connects respectively to 8 digital IO and the 4 ADC channels...

---

I wouldn't call leaving the jtag pads there bad. We do this with all of our products - no sense deliberately having a final PCB different to what you develop and test on.

As for these other pads, I don't see why they got left but I'm not going to argue about it :-)

- Pauli

Quote:

---

it appears to have a 7-segment display on the bottom line, and dot matrix on the top line. Is this possible? If so, it seems best of both worlds: segmented for numerals, matrix for alpha.

---

You were right, if the dot matrix would be in the bottom and 7-segment on the top. Reason is given here, especially in the lower part of this post.

Quote:

---

Yes, it displays the mantissa on the left side of the display and the exponent on the right side of the display.

That's exactly how the HP35, HP45, HP55, HP65, HP67....even the HP12c do it.

So, consistency! :-)

---

Consistency is not always a good thing!

That's one aspect of many HP calcs I've never liked.

My 20S and RPL calcs like the 28S/48S don't do that, so this calc is a step back IMHO.

Would be nice if it was user adjustable!

Dave.

Quote:

---

My 20S and RPL calcs like the 28S/48S don't (display exponents right-justified), so this calc is a step back IMHO.

Would be nice if it was user adjustable!

---

Hmm, if an area of the display is designated for the exponent, why would it be user-settable?

Really, there's a method to the madness....

The older models without dot-matrix displays separated the mantissa from the exponent in the interest of clarity.

The newer models with dot-matrix displays show numbers with exponents as, e.g., "2.5E49" or "2.5E-165", as they would be represented in Fortran '77. This is clearer if the "E" is reduced in size; however, only the HP-42S as well as the HP-32SII and its successors had that refinement.

The 7-segment HP-10B, HP-20S, and HP-21S display numbers with exponents in the same manner, probably for unification of microcode with their Pioneer-series brethren. Unfortunately, they can display only a full-size "E", which looks like part of a text string adjacent to blockish 0's and 1's in the display.

-- KS

Quote:

---

Hmm, if an area of the display is designated for the exponent, why would it be user-settable?

---

I'm talking about a selectable right-justified mantissa, which I greatly prefer. I also think it is much more sensible that way when you have a fixed exponent position on the right.

Quote:

---

Really, there's a method to the madness....

The older models without dot-matrix displays separated the mantissa from the exponent in the interest of clarity.

---

Yep, perfectly good reason at the time. But there is now no reason for keeping it that way with the new fixed exponent display, apart from some crazy "it's always been that way" thinking.

It's insane to have the mantissa on one side and fixed exponent on the other side, a good 30-40mm or so away for rounded numbers like say "1.1___________-3"

Quote:

---

The newer models with dot-matrix displays show numbers with exponents as, e.g., "2.5E49" or "2.5E-165", as they would be represented in Fortran '77. This is clearer if the "E" is reduced in size; however, only the HP-42S as well as the HP-32SII and its successors had that refinement.

The 7-segment HP-10B, HP-20S, and HP-21S display numbers with exponents in the same manner, probably for unification of microcode with their Pioneer-series brethren. Unfortunately, they can display only a full-size "E", which looks like part of a text string adjacent to blockish 0's and 1's in the display.

---

Yep, I dislike the big "E" a lot, not at all obvious at glance. Would have been much better to use an underscore or something.

Dave.

Edited: 13 June 2008, 9:45 a.m.

Quote:

---

The 20b is very attractive if you can live without the solver or if you really would use some of the added statistical and mathematical functions.

The solver is a very big deal for some of us. :-)

---

IMHO, HP 20b can do the job well in most cases, no matter you are a student or professional. The solver is useful if you use it to deal with complicated problems. My 17bII+ solver contains different versions of Black-Scholes option pricing formula only and not frequently use in daily work and studies.

Having a user-accessable solver is nice, but not an essential feature.

I wonder that the 20b indicates that HP has created a new architecture of future calculator design. With the 20b architechture, I hope HP could develop a "Real" successor of high end financial calculator (e.g.17bII (and bII+) and 19BII. The 20b's CPU is too powerful.

I think I will buy 20b as a replacement of my 12c platinum 25th anniversary edition for using it in exam. My 12cp's "ENTER" key was broken and dislocated from the calculator after it was dropped on the floor during the exam.